Research on Neural Network Construction Method Based on Approximate Computational Test Data

In some applications of the power grid, there are problems that the volume of real data is small and the security of real data is difficult to guarantee, which poses a challenge to the data governance model. This paper proposes a parallel convolutional neural network structure based on approximate calculation of test data, constructs test data through approximate calculation, and uses parallel convolutional neural network structure to learn the corresponding data model, which can solve the problems of data resources, computing resources and problems in data governance. Calculate the cost problem. Experiments based on existing data sets show the unique advantages of this network structure for approximately computing test data.